Analysis
Grammarly’s Parent Now Owns Your AI Detector. That Should Worry Students
Runs Detection Drama, stress-testing AI detectors and humanizers; publishes Words At Scale to 26,000+ subscribers.
Published June 26, 2026
The company that sells the writing assistant students get flagged for using just bought one of the detectors doing the flagging. Putting the accuser and the accused under one roof is good for shareholders and bad for anyone whose grade depends on the result.
On June 23, Superhuman acquired GPTZero, the AI-detection startup Princeton graduate Edward Tian first built as a senior thesis. Terms were undisclosed. Superhuman is the company Grammarly became after it bought the email client Superhuman last year and rebranded under that name. Same parent, new label.
Here is why that matters if you are a student. Grammarly is one of the tools most often blamed for tripping AI detectors in the first place. Run your own essay through it, accept a few rewrites, and Turnitin’s AI score can climb for reasons that have nothing to do with cheating. Now the parent of that editor owns a detector with more than 19 million registered users. One company profits when you polish your writing, then profits again when a machine decides the polish looks artificial.
Superhuman already shipped its own AI detector inside its product. Its stated reason for buying a second one was blunt: “two AI detectors are better than one.” For a business chasing an “authenticity layer,” maybe. For a 19-year-old whose essay got flagged after running it through Grammarly, the merger reads less like a safety feature and more like a closed loop.
The fair counter-argument is that consolidation could make detection better. GPTZero keeps its full 30-person team, co-founder Alex Cui included, and Superhuman says it will run as a standalone product. Deeper pockets could fund the accuracy work these tools badly need — detectors still flag a majority of non-native English writers as machines.
But “standalone for now” is a promise with a short shelf life, and it dodges the real problem. The conflict here is structural, not a matter of intentions. When the firm selling you the writing help also owns the judge of whether your writing is too clean, the incentive to keep both meters running is permanent. Independence is the only thing that makes a detector’s verdict worth anything, and an acquisition is the precise moment it gets harder to claim.
Tian built GPTZero in a dorm to answer a genuine question: can we tell human writing from a machine’s? Three years later the answer is being folded into the same company that sells the machine-assisted writing. That is not a detection story anymore. It is a story about who gets to grade you — and who they work for.
Vlad Ivanov
Runs Detection Drama, stress-testing AI detectors and humanizers against real student writing, and publishes the Words At Scale newsletter to 26,000+ subscribers. Connect on LinkedIn.
